Applied Mathematics E-Notes, 5(2005), 269-274 c ISSN 1607-2510

Available free at mirror sites of http://www.math.nthu.edu.tw/∼amen/

MINIMAL OPERATION TIME OF ENERGY

DEVICES∗
†
Andreas Boukas
Received 5 May 2005

Abstract

We consider the problem of determining the minimal time for which an energy
supply source should operate in order to supply a system with a desired amount
of energy in finite time.

1 Introduction
While boiling water or any other liquid, most of us have noticed that the heater can be
switched off at an intuitively chosen time and the liquid will still reach its boiling point
no long after the heater is switched off. A natural question arises: how can switch-off
time be chosen in an optimal way so that electrical energy will not be wasted? In other
words what is the earliest time at which the heater can be turned off while still reaching
the liquid’s boiling point in a finite time? A general formulation of the problem is as
follows: Let D be a device that supplies energy to a system S through a supply line.
Let E (t) be the energy supply rate. We assume that D can be switched on and off and
that it continues to supply energy, at a decreasing rate, for some time after it has been
switched off. Question: What is the minimum switch-off time of D (corresponding to
the minimal operational time of D) in order to transfer to S a total amount of energy
Q (where Q > 0 is given)? The device D can also be viewed as a control mechanism for
bringing the system S from an energy level E1 to a higher energy level E2 in finite time
while operating for the minimum time possible. Clearly, the solution of this problem
can have a lot of applications, both civilian (e.g. energy conservation) and military
(e.g. minimizing detection risk).

2 Examples
EXAMPLE 1 (Exponential Model). We consider a simple example of an energy device
supplying energy to a system at time t ≥ 0 at a rate (which in what follows we consider
∗ Mathematics Subject Classifications: 80M50, 26A06, 60H10.
† American College of Greece, Aghia Paraskevi 15342, Athens, Greece

269
270 Minimal Operation Time of Energy Devices

to include the rate at which the transferred energy is possibly radiating from the system
and/or the supply line) given by
⎧ at
⎨ e −1 if 0 ≤ t ≤ t0
E (t) = ea t 0 − 1 if t0 ≤ t ≤ t1
⎩ ea t0 −1
1−e−b T
e−b (t−t1 ) − e−b T if t ≥ t1

where a and b are positive real numbers characteristic of the source but also depending
on the environment, t0 is the time at which the energy supply rate is at its peak, and
t1 is the switch-off time. We assume that the optimal switch-off time is after the rate
of energy supply has been stabilized i.e. that tˆ1 ≥ t0 . We also assume that after
switching-off at time t1 ≥ t0 , the source stops transferring energy to the device at time
t1 + T , where T > 0 is independent of t1 . Let Q > 0 be the amount of energy that we
wish to transfer to the system. We assume that the transfer of this amount of energy
will occur at some time t2 ≥ t1 ≥ t0 . The energy Q could be, for example, the energy
required to bring a liquid substance to its boiling temperature, or the energy required
for complete phase transition. In the latter case Q = m Lv , where m is the mass of the
liquid and Lv is its latent heat of vaporization (cf. [2]). We require that
t2
E (s)ds = Q
0

which implies that

t0 t1 t2
E (s) ds + E (s)ds + E (s) ds = Q
0 t0 t1

or, by the definition of E (s),

t0 t1 t2
ea t0 − 1
(ea s − 1) ds + ea t0 − 1 ds + e−b (s−t1 ) − e−b T ds = Q
0 t0 t1 1 − e−b T

which implies that

s=t0 s=t2
ea s ea t0 − 1 e−b (s−t1 )
−s + ea t0 − 1 (t1 − t0 ) + − − e−b T s =Q
a s=0 1 − e−b T b s=t1

from which letting

y : = t2 − t1
1 1 ea t0 1 − ea t 0 1
L0 : = a t0 Q+ − + ea t0 t0 +
e −1 a a 1 − e−b T b
1
L1 : =
b(1 − e−b T )
1
L2 : =
1 − e−b T
A. Boukas 271

we obtain
t1 = L0 + L1 e−b y + L2 y. (1)
Notice that if y = 0 then t1 = L0 + L1 is the time the device supplies the desired
energy level Q without being switched-off. The maximum value of y is y = ymax =
(t1 + T ) − t1 = T . The optimal switch-off time tˆ1 is therefore determined from (1) by
letting y = T and it is given by
tˆ1 = L0 + L1 e−b T + L2 T.
Energy level Q is reached at time
t2 = tˆ1 + T = L0 + L1 e−b T + (L2 + 1)T.

EXAMPLE 2 (Linear Model). A simplified version of the previous model is obtained

by assuming that
⎧ a
⎨ t0 t if 0 ≤ t ≤ t0
E (t) = a if t0 ≤ t ≤ t1
⎩ a
− T (t − t1 ) + a if t ≥ t1
where t0 > 0, a > 0, and T is as in the model above. In this case
t2
E (s)ds = Q
0
implies that
t0 t1 t2
E (s) ds + E (s)ds + E (s) ds = Q
0 t0 t1

or, by the definition of E (s),

t0 t1 t2
a a
sds + ads + − (t − t1 ) + a ds = Q
0 t0 t0 t1 T
which implies that
s=t0
a s2 a s=t2
+ a (t1 − t0 ) + − (s − t1 )2 + a s =Q
t0 2 s=0 2T s=t1

and letting y = t2 − t1 we obtain

1 2 Q t0
t1 =
y −y+ + .
2T a 2
As in Example 1, substituting y by ymax = T we obtain
1 Q t0
tˆ1 = − T + +
2 a 2
which is greater than or equal to t0 if and only if Q 1
a − 2T −
t0
2 ≥ 0. Energy level Q is
reached at time
t0 Q 1
t2 = tˆ1 + T = + + T.
2 a 2
272 Minimal Operation Time of Energy Devices

3 General Description of the Optimal Switch-Off Time

The examples treated in detail in the previous section suggest the following general
theorems.
THEOREM 1. Let, for each t1 ≥ t0 ,
⎧
⎨ f (t) if 0 ≤ t ≤ t0
Et1 (t) = f (t0 ) if t0 ≤ t ≤ t1
⎩
g(t) if t ≥ t1
where f is continuous and increasing with f (0) = 0, and g is continuous and decreasing
with g(t1 ) = f (t0 ). Let F and G denote the anti-derivatives of f and g respectively. If
there exists T > 0 such that g(t1 + T ) = 0 for all t1 ≥ t0 , then the optimal switch-off
time tˆ1 ≥ t0 is the solution of
F (t0 ) + f (t0 ) (tˆ1 − t0 ) + G(tˆ1 + T ) − G(tˆ1 ) = Q.
PROOF. The condition
t2
Et1 (s) ds = Q
0
implies that
(F (t0 ) − F (0)) + f (t0 )(t1 − t0 ) + (G(t2 ) − G(t1 )) = Q
which by F (0) = 0 and the fact that for the optimal tˆ1 , t2 = tˆ1 + T implies that
F (t0 ) + f (t0 ) (tˆ1 − t0 ) + G(tˆ1 + T ) − G(tˆ1 ) = Q.
We can generalize the above theorem to the case of a heat source that supplies
energy at a possibly strictly increasing (or even arbitrary continuous) rate as follows.
THEOREM 2. Let, for each t1 ≥ t0 , the energy function Et1 (t) be an increasing
continuously differentiable function of t. The optimal switch-off time tˆ1 and the as-
sociated time t2 at which the energy level reaches Q are determined by the system of
equations
Etˆ1 (t2 ) = Q, Etˆ1 (t2 ) = 0.

PROOF. It is clear that the optimal switch-off time tˆ1 makes full use of the energy
source in the sense that the desired energy amount Q is supplied at the moment when
the source dies out i.e.e when Etˆ (t2 ) = 0. For such a t2 ,
1

t2
Etˆ1 (s) ds = Q
0
implies
tˆ1 t2
Etˆ1 (s)ds + Etˆ1 (s)ds = Q
0 tˆ1

i.e.e Etˆ1 (tˆ1 ) − Etˆ1 (0) + Etˆ1 (t2 ) − Etˆ1 (tˆ1 ) = Q and since Etˆ1 (0) = 0, we obtain Etˆ1 (t2 ) =
Q.
A. Boukas 273

4 Noisy Supply Line

The energy supply rate functions φt1 (t) := E (t) used in Examples 1 and 2 above, can
be viewed on each of the intervals [0, t0 ], [t0 , t1 ] and [t1 , t2 ], as solutions of ordinary
differential equations of the form
dφt1 (t) = (c1 φt1 (t) + c2 )dt (2)
for appropriate constants c1 , c2 (depending on the interval) and initial condition φt1 (0) =
0, with the different branches of φt1 (t) tied up at t0 and t1 . It could be the case how-
ever, that the supply line connecting the energy source D with the system S is affected
by noise, appearing in (2) in the form of additive noise
dφt1 (t) = (c1 φt1 (t) + c2 )dt + (c3 φt1 (t) + c4 ) dB(t) (3)
or equivalently
t t
φt1 (t) = (c1 φt1 (s) + c2 ) ds + (c3 φt1 (s) + c4 ) dB(s) (4)
0 0

where c1 , c2 , c3 , c4 are constants, B(s) is one dimensional Brownian motion and the
stochastic integral on the right hand side of (4) is in the sense of Itô (cf. [1]). In
that case φt1 (t) is actually a stochastic process φt1 (t, ω) and the problem of finding the
optimal switch-off time tˆ1 now amounts to finding the first t1 for which
t2
µt1 (s)ds = Q
0

for some finite t2 ≥ t1 , where µt1 (t) denotes the mathematical expectation of φt1 (t, ω)
. Since the mathematical expectation of an Itô stochastic integral with respect to
Brownian motion is equal to zero, (4) implies upon taking the expectation of both
sides that
t
µt1 (t) = (c1 µt1 (s) + c2 ) ds
0

which can be solved explicitly and yields the mean energy supply rate
c2 c1 t
µt1 (t) = e −1
c1
where on each interval the constants c1 and c2 are determined by using the initial and
tying up conditions. Thus we are reduced to the deterministic problem considered in
the examples of Section 2, but this time for the mean energy supply function. The
method extends directly to the case when equations (2) and (3) are replaced by the
more general equations
dφt1 (t) = f (t, φt1 (t)) dt
and
dφt1 (t) = f (t, φt1 (t)) dt + g(t, φt1 (t))dB(t)
respectively.
274 Minimal Operation Time of Energy Devices

References
[1] B. Oksendal, Stochastic Differential Equations, 2nd edition, Springer-Verlag 1989.
[2] J. S. Faughn, R. A. Serway, C. Vuille and C. A. Bennett, College Physics, 7th
edition, Thomson Brooks/Cole 2006.